I think a better (separate?) interface could also speed up things significantly - like “gamifying” it with arrow left/right and having the next observation showing up instantly. I have been playing with that idea for detecting bird bands; like making a small webapp to quickly assess all observations of a species in a region: The photos show up, use keyboard shortcuts to set observation fields (e.g., a → Banded = yes, s → Banded = no; d → Banded = Not possible to assess), and on click, the next observation is shown. That saves a lot of clicks and time
This is already possible, see e.g.
https://forum.inaturalist.org/t/using-identify-to-annotate-observations/1417
2 Likes
Nice, thanks, I didn’t know!
1 Like
It’s a good sentiment but this thread is almost 7 years old and the situation is slightly better in percentage terms, and abysmal in absolute terms. I had a look at the situation with Life Stages for insects, which is what got me started on this topic.
Starting with the relatively good news - looking at global Lepidoptera, the rate of adding Life Stage annotations has increased to a cumulative 61% or so. However, that means that 39% of the 32 million lep observations have no Life Stage annotation. Rounded, that is 12.5 million unannotated observations. We are not likely to tackle that manually. And Leps are the bright spot, the aggregate ratio for insects as a class is almost exactly the inverse, with 61% of observations having no Life Stage annotation. That is 48 million unannotated observations. The numbers are huge and only grow.
The overwhelming majority of users don’t read helpful threads. We need to nudge them with technology.
There have been several rebuttals about the need for training data. I agree that we do need training data, and there are now millions of annotated observations. This seems like CV in the early days, you start with the species with rich data and move on. There are almost 2.9 million Lepidoptera larvae observations, that is more adults than we had for the early CV versions.
I largely agree with Paul’s comments, especially that we should be selective about the taxa involved. But on the other hand, the CV definitely can identify larva to species already. All we are asking for is more granular classification of those ID suggestions.
1 Like
Those numbers are really eye-opening. 48 million unannotated insect observations is a mountain we’re never going to climb manually. Using CV to at least suggest life stages where the data is already rich (like Lepidoptera) seems like a logical next step to help chip away at that backlog.
And now the I see the ongoing requests for help with assigning annotations in this thread:
https://forum.inaturalist.org/t/annotations-happy-to-help/74137
I wonder if all those hours couldn’t be better spent allowing users to connect with nature by finding a technical solution to the workflow?
if someone has time, and an internet connection, and enjoys annotating, while learning as they go - why not ? That is where the other thread started and it has support. Not every iNatter wants to add whatever to the existing Needs ID mountain. It is good to have another pair of eyes on an obs - wait - that’s not …
Also the milestones thread has 2 active identifiers who have no obs so far this year.
3 Likes
I agree to some extent. However, speaking personally, I’m very connected to nature and am outdoors a lot, but it’s currently winter and dark outside and anyway, I like having the opportunity to learn something new.
4 Likes
I hear you, but given the mountain of IDs needed, doesn’t it seem excessive to ask people to annotate if the software could enable it while uploading?
I want to emphasize that I respect all those contributions. I have almost twice as many annotations as IDs, and I believe that given limited time and volunteer hours, automating as much as possible is preferred.
1 Like
Would there be a way to opt out?
1 Like